Causal reasoning
Causal reasoning is the ability to identify relationships between causes - events or forces in the environment - and the effects they produce. Humans and some other animals have the ability not only to understand causality, but also to use this information to improve decision making and to make inferences about past and future events.Garcia-Retamero, R., Wallin, A., & Dieckmann, A. (2007). "Does causal knowledge help us to be faster and more frugal in our decisions?" Memory & Cognition, 35, 1399–1409. An invariant that guides human reasoning and learning about events is causality.Sloman, S. A. & Lagando, D.A. (2003). "Causal invariance in reasoning and learning." The Psychology of Learning and Motivation, 44, 287-325. Causal considerations are integral in how people reason about their environment.Sloman, S. A. (2005). Causal models. New York, NY: Oxford University Press. Humans use causal cues and their related effects to make decisions efficiently, to make predictions about the future circumstances of our environment and to fully understand mechanisms leading to change.Hagmayer, Y., Sloman, S. A., Lagnado, D. A., & Waldmann, M. R. (2007). "Causal reasoning through intervention." In A. Gopnik & L. Schulz (Eds.), Causal learning: Psychology, philosophy, and computation (pp. 86–100). Oxford, England: Oxford University Press. Understanding Cause and Effect Causal relationships are often understood as a transfer of some sort of force.Ahn, W.-K., & Kalish, C. W. (2000). "The role of mechanism in causal reasoning." In F. C. Keil & R. A. Wilson (Eds.) Explanation and Cognition. (pp. 199-225). Cambridge, MA: The MIT Press. If A is caused by B, then A must transmit some sort of force or causal power to B which results in this particular effect. Causal relationships suggest a change happening over time—the cause and effect are temporally related such that the cause precedes the outcome. Causality can also be inferred in the absence of a force, although this is a less typical definition.Wolff, P., Barbey, A. K., & Hausknecht, M. (2010). "For want of a nail: How absences cause events." Journal of Experimental Psychology: General, 139, 191-221. For example, a cause can be the removal or stopping of an event, like removing a support from a structure would cause a collapse; or a cause could be the non-occurrence of an event, like a lack precipitation causing wilted plants. Although it may seem difficult to reconcile the idea of a non-event actually causing something, it is still possible for this sort of relationship to be consistent with typical causality theories. This sort of causation can be inferred as the force that is preventing the effect being removed by another force. So there is still some causal agent, it is just further removed from the ultimate effect. Humans can reason about many topics with the aid of causal understanding. Examples include social situations, counterfactual situations, and mathematics. Understanding these topics depends on the ability to understand cause and effect. People must be able to reason about the causes of others’ behavior in order to understand their intentions and act appropriately with them. People must also understand the likely effects of our own actions. Counterfactual arguments are presented to individuals in many situations—humans are predisposed to think about “what might have been”, even when that argument has no bearing on the current situation. An understanding of cause and effect is crucial for these sorts of arguments, as people must appreciate what caused our current situation and what circumstances could change to result in a different state of the world. Finally, mathematics and other logical disciplines require an understanding of cause and effect and the logical rules that must follow when certain actions are taken. Mathematics in particular has well defined rules for logical proofs, for example, in which certain effects must always follow particular causes. Causality is also directly related to mechanism.Keil, F. C. (2006). "Explanation and understanding." Annual Review in Psychology, 57, 227-254. The understanding of a mechanistic relationship allows an individual to infer causality. For example, if you know the details of how a bicycle operates, you understand that the pedals cause the chain to turn gears in the rear of the bicycle and move it forward. In many circumstances, people confound their understanding of causal relationships with their understanding of mechanism. For example, they may know that an air conditioner causes a home to cool down, and based on this knowledge assume that they understand the mechanistic properties of an air conditioning system, when in fact they do not. An understanding of causality does not necessarily imply an understanding of mechanism. Cause and effect relationships help people define the categories of objects, for example, understanding that a feature that identifies a category is causally related to the properties of the members of that category allows us to more concretely describe those categories.Rehder, B. (2003). "Categorization as causal reasoning." Cognitive Science, 27, 709-748. For example, one might understand that “wings” is one key feature of the category members “birds”, and this feature is causally interconnected to another inherent feature of that group, which is the ability to fly. This discussion of birds exemplifies something called the Causal Model Theory of categories and causality, which suggests that people’s intuitive theories about a category depend on both the observable features inherent to that category as well as causal mechanisms. How Humans Infer Cause and Effect Humans are predisposed to understand cause and effect, and use many strategies to make inferences about causes and effects bi-directionally. In particular, people use temporal cues to understand causality.Lagnado, D. A., & Sloman, S. A. (2006). "Time as a guide to cause." Journal of Experimental Psychology: Learning, Memory, and Cognition, 32, 451–460. When observing an event, people tend to assume that things directly preceding that event are candidates for causes of it, and things directly following that event are effects from it.Cheng, P. W., & Novick, L. R. (1990). "A probabilistic contrast model of causal induction." Journal of Personality and Social Psychology, 58, 545-567. Co-occurrence of movement and spatial relationships are another way in which people infer cause and effect. If objects move together or in such a way as to appear that one object is initiating the movement of another, we infer causality from that relationship.Scholl, B. J. & Tremoulet, P. D. (2000). "Perceptual causality and animacy." Trends in Cognitive Sciences, 4, 299-309. In fact, in many cases we can also infer animacy from such relationships. For example, if one object moves about a computer screen and another object moves slightly behind it, we may infer that the second object is “chasing” the first, despite the fact that these objects can have no motivation of their own. This is the Michotte effect, attributed to the researcher who first discovered it. This effect also occurs if one object moves towards a second objects position, stops at that location, and the second object then rapidly moves along the same trajectory. In this case, the first object is seen as causing the action of the second object. This illusion is particularly interesting because there is nothing causal actually occurring between these computer animations at all—they operate completely independently, yet our bias to understand causality influences us to see the two objects as animate and having a causal relationship. Basic causal processing can be activated almost automatically. We are very well practiced in understanding cause and effect, and this practice results in a low-effort process when making this sorts of inference.Corrigan, R. & Denton, P. (1996). "Causal understanding as a developmental primitive." Developmental Review, 16, 162-202. However, while we are very good at inferring cause and effect, we do not always understand the mechanisms underlying that causality. In fact, causality has been described as a “cognitive illusion”.Sawa, K. (2009). "Predictive behavior and causal learning in animals and humans." Japanese Psychological Research, 51, 222-233. Much of our understanding of cause and effect is based on associations, without a true understanding of how events are really related to one another.This lack of understanding is referred to the as the Illusion of Explanatory Depth. People may believe they have an understanding of mechanistic relationships because they understand that one event causes another, but when asked to explain that relationship, they fail. Types of Causal Relationships There are many different types of causal models we develop as a result of observing causal relationships in the world. In particular, there are common cause relationships, common effect relationships, causal chains, and causal homeostasis. Common cause relationships involve a single cause resulting in multiple effects. For example: A virus could be an example of a single cause that results in multiple related effects like a fever, headache, and nausea. Common effect relationships involve multiple causes converging on one effect. For example: An increase in government spending could be an example of a single effect with multiple causes. The increased spending could be caused by high unemployment, increases in currency value, or civil unrest for example. These causes all converge to produce this effect. Causal chains involve a single cause instigating an effect which in turn instigates another effect and so on. For example: The “butterfly effect” is one example of a causal chain. In this figurative example, a butterfly may flap its wings, which can result in disturbances in air currents, which can cause a hurricane, which can in turn cause damage to a community. A more realistic example could be poor sleep leading to fatigue, which in turn leads to clumsy coordination. Causal homeostasis involves causal relationships that endure as a stable cycle or reinforcing mechanism. For example: The presence of feathers, hollow bones, high metabolic rate, and flight may all reinforce each other in birds in this way, with continuous adaptation to the whole cycle rather than one instance particularly beginning the causal relationship. Types of Reasoning about Cause and Effect While basic causal understanding can be automatic, there are more complex situations in which advanced reasoning may be necessary. There are several different ways we can apply reasoning to infer causality, including: Deduction Deduction implies a general rule. If some event occurs, there is a guaranteed conclusion. In deductive reasoning, one can infer an outcome based on the presence of other arguments, and these arguments can be used to determine a cause and effect relationship. For example, one might say: Earthquakes cause damage to property and people. Damage is a negative outcome. Therefore earthquakes cause negative outcomes. Based on the first two statements in this example, the third is a logical conclusion with no ambiguity. Induction Induction includes any inference made under uncertainty. In this case the conclusion is likely but not guaranteed. In this way, induction can be used to speculate about causality, but true causal understanding is not likely to come from this type of reasoning. For example, when using inductive reasoning to infer causality one might say: All evidence to this point suggests that genetic mutations produce cancerous tumors. Therefore, it is likely that all cancer is caused by genetic mutations. In this example, the first statement is plausible, and the second is a conclusion that is likely but not guaranteed because more evidence may still come to light about different causes. Abduction In this case the premises do not guarantee conclusion. Abductive reasoning moves from data description to a hypothesis without a necessarily intransient relationship between the cause and effect. This is a less traditional form of reasoning for these circumstances. For example, one might observe a door opening in a room, and abduce that the wind opened this door. There may be many causal factors that could have resulted in that outcome, but the hypothesis of the wind acting as a cause is reasonable. We are inclined to search for a single causal explanation for effects in our environment in order to lessen ambiguity. Perspectives on Causal Reasoning There are several theories and models of how humans reason about causality. Dependency Model The dependency model (or class of interrelated models) asserts that effects are contingent upon causes. Cause and effects have a probabilistic relationship in this sort of model. For example, cancer is more likely in the presence of smoking, so cancer is assumed to be caused by smoking. Covariation Model The covariation or regularity model suggests that humans understand the relationships between causes and effects by virtue of their co-occurrence and how changes in a cause result in changes in an effect.Cheng, P. W. (1997). "From covariation to causation: A causal power theory." Psychological Review, 104, 367–405. This is a type of dependency model. This model would propose that we notice correlations between events, observe concurrent changes to these events, and from this covariation infer causality. Critically, this co-variation must occur with one of the events imparting causal power. The rooster crowing may always coincide with the sun rising but one does not infer the rooster causes the sun to rise because it does not exert causal power Mechanism Model This model suggests that causes and effects are related by a mechanistic relationship. In this situation, there is proposed to be a basic process underlying the cause and effect relationship. For example if a sneeze gets someone sick, it is understood that there are germs transmitted by the sneeze—they are the mechanism by which illness is transferred. Dynamics Model This model of causal representationWolff, P. (2007). Representing causation. Journal of Experimental Psychology: General, 136, 82-111. suggests that causes are represented by a pattern of forces and a position vector. What this means is that some physical force in the world (gravity, momentum, chemical forces, etc.) act between events or objects to produce some sort of an end state. For example, a rubber ring on the bottom of a drink coaster may cause it to stay in place on the table. There is no action in the typical sense of the word occurring in this example, but it is causal because of the physical forces resulting in an end state (lack of movement). Development of Causal Reasoning in Humans Children develop the ability to understand causality and make inferences based on cause and effect from a very young age. Some research suggests children as young as 8-months old can understand cause and effect.Sobel, D. M., & Kirkham, N. Z. (2006). Blickets and babies: The development of causal reasoning in toddlers and infants. Developmental Psychology, 42, 1103-1115. Understanding mechanism and understanding causality go hand in hand; children need to understand that there are causes and effects in the world in order to understand how mechanisms can operate, and this knowledge in turn allows them to understand specific causal relationships. Children begin to ask “why” at a very early age, and do so in order to understand mechanism and in turn causality. Children’s first instance of a “why” question often coincides with their first attempts to explain things, explanations which are often causal in nature and occur within the first year after the child acquires language. Children ask “why” expressly for the purpose of understanding mechanism and causality, and persist in asking such questions until they get mechanistic or causal information as a response.Keil, F. C. (2012). Running on empty? How folk science gets by with less. Current Directions in Psychological Science, 21, 329-334. The ability to understand and reason about causality at a young age allows children to develop naïve theories about many topics. For example causality helps children to learn about physics, language and concepts, and the behavior of others. For example, a child may develop naïve theories of gravity based on the observation that something must cause dropped objects to fall to the ground. They may develop theories of language and conceptual representations because of their understanding that specific features of objects cause people to apply labels in a consistent way. Or they may develop theories about the intentions of others based on the observation that something must cause another person to act in the ways they do. There is a clear developmental pattern of the types of causal understandings children can have at various ages. Some levels of understanding about causality emerge in infancy, other levels emerge in childhood, while others still emerge later in adulthood or not at all. Infants have an understanding of causal power. They know that certain causes can result in particular effects. They also understand causal relevancy. They understand that certain properties are more relevant to particular relationships, and can track these properties in relation to the causal mechanisms in question. Young children, from late infancy to early childhood, understand functional relations. They can understand that a particular property or component of a mechanism can serve a certain function. They also understand causal density, which means that they understand how different causes can interact and connect in a complex way. Older children and adults continue to develop an understanding of mechanistic fragments throughout the lifespan. They begin to understand the concrete components of a working system in an isolated way, although full mechanistic details of a system do not emerge until adulthood and sometimes not even at that point. It requires a level of expertise to be able to fully describe and manipulate the full understanding of a mechanistic system. One theory on the development of causal reasoning, for instance, is that of Jean Piaget, who defined a preoperational stage, concrete operational stage, and formal operational stage of development. Causal Reasoning Across Cultures Causal attributions have been shown to be dissimilar among different cultures in many domains: Causal Attributions Yan and GaierYan, W. & Gaier, E. L. (1994). Causal attributions for college success and failure: An Asian-American comparison. Journal of Cross-Cultural Psychology, 25, 146-158. investigated different causal attributions of college success and failure between two different groups of college participants: Americans and Asians. The Asian student group was made up of Chinese, Korean, Japanese, and Southeast Asian students, and performance on this task was very similar across all four nationalities. The students were asked to make judgments about someone else's successes and failures in schoolwork, and whether those outcomes were attributable to innate ability or to expended effort. American participants were much more likely to attribute academic achievement to ability than Asian participants were. Americans also tended to rate success as being attributable to effort, whereas failure was not perceived as being a result of a lack of effort. Asian students did not show this pattern. Comparisons between Western and Eastern children and adults suggest that there are differences between the cultures in the causality attributable to particular illnesses.Nguyen, S. P. & Rosengren, K. S. (2004). Causal reasoning about illness: A comparison between European and Vietnamese-American children. Journal of Cognition and Culture, 4, 51-78. While reading stories of illnesses and making inferences about the causes of those illnesses, both groups showed an understanding of biological causes of most illnesses. However, both Western and Eastern children and Eastern adults also attributed some illnesses and their remedies to magical causes. Western adults did not make those attributions. Between these cultures there are somewhat different understandings of particular illnesses, how they are contracted, and how they are remedied. Causal Motivations People from individual or collectivist cultures may make different attributions as to the origins and motivations of movement on a small scale among animated objects, or what would cause movement within a group of animated objects.,.Ng, S. H. & Zhu, Y. (2001) Attributing causality and remembering events in individual- and group-acting situations: A Beijing, Hong Kong, and Wellington comparison. Asian Journal of Social Psychology, 4, 39-52. Participants from the UK or China or Hong Kong were shown videos of animated fish moving about a computer screen. The videos depicted one central actor fish moving either towards or away from a group of fish, and the critical task was included in judgments participants made. They were asked to choose what statement represented the relationship among the fish. This relationship could be either internally motivated, for example the central fish was swimming because it was looking for food, or externally motivated, for example the central fish was swimming because it was charmed by the other group of fish and wanted to join them. Another set of videos suggested that the group of fish was the predominant acting agent, while the individual fish was being acted upon. These different videos provide an opportunity to determine whether group or individual action is the preferred motivating force among different cultures. Results suggested that Asian participants preferred descriptions and situations in which the group was the central focus and causal agent, while the Western participants preferred the situations in which the individual was the causal agent. This preference rating was a self-report measure from the participants. These effects also extended to memory processes—participants from the collectivist group had better memory for the situations in which the group was primary causing the actions. These results suggest that members of individualistic cultures are more responsive to independent agents, whereas members of collectivist cultures are more responsive when groups guide an individual’s action. Causal Reasoning in Non-human Animals Causal reasoning is important to humans but it is not unique to humans. Animals are often able to use causal information as strong cues for survival. Specifically, rats are able to generalize causal cues in order to gain new food rewards. So, for example, they can generalize beyond simply “do x and get y”, to “do something related to x and get y”. Animals like rats can learn the actual mechanisms required for a rewarding effect, and then reason about what sorts of causes could elicit that reward in order to earn it (Sawa, 2009). (Corvus moneduloides)]] New Caledonian crows are a species of animal that have been studied for their ability to reason about causal events.Hunt, Gavin R. (January 1996). "Manufacture and use of hook-tools by New Caledonian crows". Nature 379: 249–251. These birds are a very intelligent species, using tools even in a way that chimpanzees cannot—they can also even make their own complex tools to solve problems such as food out of reach. Experimental work with this species of crows suggests that they can even understand hidden causes in a way that was previously believed to be uniquely human.Taylor, A. H., Miller, R., & Gray, R. D. (2012). New Caledonian crows reason about hidden causal agents. Proceedings of the National Academy of Sciences, 109, 16389-16391. Two experiments suggest such an inference. Crows were placed in a confined area with food in a tube that was inaccessible to the crow without some effort on the crow’s part. In one experiment, a human walked into the enclosure behind a curtain, and started to move a stick around the area of the food tube through a hole in the curtain. After the human left, the crow confidently moved toward the food area and retrieved the reward quickly. The crow understood that human was the cause of the moving stick, even though the human was not visible when the stick was moving. Therefore, when the human walked away, the crow understood that the stick would not be moving on its own. In the second experiment, the stick moved from behind the curtain in the same way as the first experiment, but there was no human entering or exiting the enclosure. In this case, when the crow moved toward the food, it was with trepidation and it nervously looked toward the hole in the curtain, actions which slowed down its access to the reward. The authors suggest that this means the crow could understand hidden causes. In the first scenario, the crow knew that the cause of the moving stick was the human, even though the human was not actually visible while the stick was moving. In the second experiment, the crow did not know the cause of the moving stick, so therefore was not assured that the stick would not reappear in the area of the food reward while it was retrieving the food. Since the human could not be inferred as the cause of the movement, the crows understood that the stick could reappear at any time. These animals have a sophisticated understanding of causal events, even when the cause is not visible to them. References Category:Causality Category:Reasoning